DeepSeek-V3 vs Gemini 2.5 Pro
Wins 0 of 6 areas
—
Wins 5 of 6 areas
Coding · Agents · Reasoning · Long documents · Following instructions
Gemini 2.5 Pro is the stronger all-rounder.DeepSeek-V3 is cheaper.
Scores updated · 45 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software04Gemini 2.5 Pro4 of 4 tests
- ReasoningHard problems that need careful thinking04Gemini 2.5 Pro4 of 4 tests
- Long documentsFinding answers in very long texts02Gemini 2.5 Pro2 of 2 tests
- Following instructionsDoing exactly what it is asked02Gemini 2.5 Pro2 of 2 tests
- AgentsCarrying out multi-step tasks on its own01Gemini 2.5 Pro1 of 3 tests · 2 ties
- FactsGetting facts right instead of making them up11Even1 each · 1 tie
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
DeepSeek-V3 costs 90% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Gemini 2.5 Pro pulls ahead
- Common-sense trick questionsSimpleBench+32.7points ahead
- Reasons across sets of long documentsAA-LCR+28.3points ahead
- Keeps track of context across a multi-turn chatMulti-Challenge+22.2points ahead
Where DeepSeek-V3 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+5points ahead
Every test, side by side
All 45 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemini 2.5 Pro
- SWE-bench VerifiedGemini 2.5 Pro by 21.24263.2+21.2
- Terminal-Bench 2.1Gemini 2.5 Pro by 11.616.928.5+11.6
- Terminal-Bench HardGemini 2.5 Pro by 11.315.226.5+11.3
- SciCodeGemini 2.5 Pro by 7.33946.3+7.3
AgentsGemini 2.5 Pro
- τ-Bench V3 · BankingGemini 2.5 Pro by 54.79.7+5
- GDPValtie00tie
- Terminal-Bench 4.0tie00tie
ReasoningGemini 2.5 Pro
- SimpleBenchGemini 2.5 Pro by 32.718.951.6+32.7
- GPQA DiamondGemini 2.5 Pro by 18.165.583.6+18.1
- Humanity's Last ExamGemini 2.5 Pro by 13.34.718+13.3
- CritPtGemini 2.5 Pro by 2.602.6+2.6
FactsEven
- AA-Omniscience · AccuracyGemini 2.5 Pro by 13.625.439+13.6
- AA-Omniscience · Non-hallucinationDeepSeek-V3 by 514.19.1+5
- Vectara HHEM hallucination ratelower is bettertie6.17tie
Long documentsGemini 2.5 Pro
- AA-LCRGemini 2.5 Pro by 28.340.769+28.3
- LongBench v2Gemini 2.5 Pro by 10.148.758.8+10.1
Following instructionsGemini 2.5 Pro
- Multi-ChallengeGemini 2.5 Pro by 22.231.453.6+22.2
- IFBenchGemini 2.5 Pro by 7.74148.7+7.7
Other results27 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- AIME 2024Gemini 2.5 Pro by 52.839.292+52.8
- HMMT Feb. 2025Gemini 2.5 Pro by 51.629.280.8+51.6
- livecodebench_hardGemini 2.5 Pro by 45.214.259.4+45.2
- HMMT 2025Gemini 2.5 Pro by 38.227.565.7+38.2
- livecodebench_mediumGemini 2.5 Pro by 37.153.590.6+37.1
- AIR-Bench 2024Gemini 2.5 Pro by 32.840.873.6+32.8
- AIME 2025Gemini 2.5 Pro by 31.751.383+31.7
- Aider-PolyglotGemini 2.5 Pro by 26.949.676.5+26.9
- SimpleQAGemini 2.5 Pro by 25.924.950.8+25.9
- LiveCodeBenchGemini 2.5 Pro by 2549.274.2+25
- AA-OmniscienceGemini 2.5 Pro by 24.4-40.7-16.4+24.4
- Artificial Analysis Coding IndexGemini 2.5 Pro by 23.72346.7+23.7
- τ²-BenchGemini 2.5 Pro by 21.532.554+21.5
- HarmBenchGemini 2.5 Pro by 15.749.765.4+15.7
- livecodebench_easyGemini 2.5 Pro by 15.583.398.8+15.5
- MMLU-ProGemini 2.5 Pro by 10.175.986+10.1
- MATH-500 (EM)Gemini 2.5 Pro by 8.690.298.8+8.6
- ZebraLogicGemini 2.5 Pro by 7.68491.6+7.6
- τ²-Bench Telecom (AA run)Gemini 2.5 Pro by 747.154.1+7
- AA IntelligenceGemini 2.5 Pro by 5.39.715+5.3
- vectara_avg_summary_lengthGemini 2.5 Pro by 24.7 rating points81.7106.4+24.7 rating
- anthropic_red_teamGemini 2.5 Pro by 2.497.199.5+2.4
- simple_safety_testsGemini 2.5 Pro by 1.795.397+1.7
- vectara_answer_rateGemini 2.5 Pro by 1.697.599.1+1.6
- XSTestGemini 2.5 Pro by 1.697.198.7+1.6
- vectara_factual_consistencytie93.993tie
- bbqtie96.796.4tie
Questions people ask
Which is better, DeepSeek-V3 or Gemini 2.5 Pro?
Gemini 2.5 Pro wins five of the six areas where both have results: coding, agents, reasoning, long documents and following instructions. DeepSeek-V3 wins none, but costs 90% less. They are level on facts.
Which is better for coding?
Gemini 2.5 Pro. It wins 4 of the 4 coding tests both models report; DeepSeek-V3 wins none.
Which is cheaper?
DeepSeek-V3 costs $0.24 per million input tokens and $0.90 per million output tokens; Gemini 2.5 Pro costs $1.25 and $10.00. That makes DeepSeek-V3 about 90% cheaper for the same work.
How do you compare the two?
We use the 45 benchmark tests both models have published scores on. The verdict counts the 18 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 27 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.